Sona

Sona is an intelligent recommendation service for book lovers, developed by a team of enthusiasts at the intersection of philology and data science. Its main value lies in deeply personalized literature selection that goes beyond standard "similar books" algorithms, focusing on the user's current mood, emotional needs, and subtle reading preferences. The service is built on a hybrid model combining large language models (LLMs) for understanding context and mood with collaborative filtering that analyzes behavior patterns of millions of readers.

Key features: Sona allows users to describe their state or desires in free text form — for example, "I want a book that smells like rain and loneliness" or "I'm looking for a story about overcoming with a Middle-earth atmosphere." The system analyzes these queries, considering not only genre and plot but also narrative pace, prose density, emotional tone, and even author's style. The service creates personalized selections with detailed explanations of choices, maintains recommendation history, and allows refining tastes through interactive dialogue, continuously learning from feedback.

Sona's uniqueness lies in its psycholinguistic approach and rejection of template categories in favor of semantic analysis of users' textual descriptions. Unlike most aggregators and recommendation systems of major retailers, Sona is not tied to a specific platform or store, and its algorithm considers "mood-based" metadata extracted from reviews and literary analysis. The service is available as a web application and mobile app for iOS, offering a free basic tier with limited requests and a Premium subscription for unlimited deep recommendations and access to exclusive thematic collections.

Ideal for avid readers experiencing "genre fatigue," book bloggers seeking fresh angles for reviews, as well as for bibliotherapists and psychologists using literature in their work. The service is indispensable in scenarios where you need to find a book for a specific emotional state, recall a work from fragmented plot memories, or build a deeply personalized reading plan for the year.